Consider an illustrative scenario, one that is common among e-commerce businesses spending serious money on ads.
They look at their channel performance report. Google retargeting shows a 4.2x ROAS. Facebook prospecting shows 1.1x. The obvious conclusion — cut Facebook, double down on retargeting. They do it. Six months later, conversions are declining and they cannot figure out why.
What actually happened is that they starved the top of their funnel and then wondered why nobody was entering it.
Last-click attribution told them retargeting was working brilliantly. What it did not tell them was that retargeting was converting customers that Facebook prospecting had already brought into the consideration stage. Kill the prospecting, and eventually the retargeting has nobody left to convert.
This is one of the most common budget allocation mistakes in e-commerce marketing.
What last-click attribution actually measures
Last-click attribution gives 100% of the credit for a conversion to the final touchpoint before the sale. If a customer clicked a Google retargeting ad and then bought, Google retargeting gets the credit. Everything that happened before that click — the Instagram post that introduced the brand, the email sequence that built consideration, the blog article that answered their question — gets nothing.
It became the default in most ad platforms because it is simple to implement and easy to explain. If you clicked it last, you caused it. Done.
The problem is that it measures the last step of a customer journey, not the journey itself. And for almost every customer at every price point above impulse-buy territory, the journey is the thing that matters.
What a real customer journey looks like
For anything above an impulse purchase, customers rarely buy after a single interaction. They typically meet a brand several times, across several channels, before the final click. A simplified journey for a mid-priced e-commerce product often looks like this:
- Discovery: paid social (Meta prospecting) or organic search
- Consideration: email sequence, organic social, direct visits
- Intent: branded search, retargeting
- Conversion: retargeting click or direct
Under last-click attribution, retargeting and branded search, the channels at the end of this journey, tend to claim most of the revenue credit. A position-based model, which weights the first and last touches more heavily than the middle ones, gives paid social and email back credit that last-click makes invisible.
The channels that look weakest in a last-click report are often doing much of the work of bringing customers in.
The specific ways last-click distorts your decisions
It makes retargeting look like a miracle and prospecting look like a waste
Retargeting converts well in last-click models because it catches customers who are already close to buying. It did not make them close to buying. Something earlier in the funnel did that. But last-click attributes the conversion to the final touch, which is usually retargeting.
The result is that businesses over-invest in retargeting until the audience pool shrinks because nobody is being added to the top of the funnel, and then wonder why retargeting ROAS is falling.
It undervalues brand awareness channels almost completely
Display advertising, organic social, influencer content, YouTube — these channels almost never show up as the last touch before conversion. Under last-click, they appear to generate zero return. So they get cut. The customer who saw three Instagram posts before searching for the brand on Google and converting is invisible in last-click data. Google branded search takes the credit.
It distorts your understanding of customer acquisition cost
If you are calculating CAC channel by channel using last-click conversions, you are undercounting the cost of acquiring customers through channels that do early-funnel work. Your paid social CAC looks high because it is taking the cost without getting the conversion credit. Your branded search CAC looks impossibly low because it is getting the credit without the cost.
Better models and when to use each
There is no universally correct attribution model. The right one depends on your product, your average customer journey length and what decisions you are trying to make. Here is how we think about it:
Linear attribution
Distributes credit equally across all touchpoints in the path. Simple, fair, easy to explain. Good starting point if you have never done multi-touch attribution before. The weakness is that it treats a first awareness touchpoint the same as a final intent touchpoint, which is not quite right either.
Time-decay attribution
Gives more credit to touchpoints closer to the conversion, less to earlier ones. Better than linear for businesses with short consideration cycles. For impulse purchases or low-ticket products, it makes sense that recent touchpoints drove more of the decision. Less useful for considered purchases where the research phase matters as much as the final click.
Position-based (U-shaped) attribution
Gives 40% to the first touch, 40% to the last touch, and distributes the remaining 20% across middle touches. Our preferred starting model for most e-commerce businesses because it recognises both the acquisition moment (first touch, which told the customer the brand existed) and the conversion moment (last touch, which closed the sale) while not completely ignoring the middle. Note that GA4 no longer offers this model in its reports, so it is typically built from exported path data in SQL or Python.
Data-driven attribution
Uses machine learning to assign credit based on the actual contribution of each touchpoint to conversion probability. It is the default reporting attribution model in Google Analytics 4. It is the most data-informed of the four, but also the hardest to explain to a non-technical stakeholder, and it is most reliable when an account has a healthy volume of conversions.
How to tell if last-click is distorting your budget decisions
Ask yourself these questions about your current channel performance data:
Are you seeing retargeting ROAS consistently above 3x while prospecting channels are below 1x? That gap is often last-click attribution working as described above, not a genuine signal that prospecting is not working.
When you cut a top-of-funnel channel, did retargeting performance decline 4–8 weeks later? This lag is the signal. If retargeting starts degrading weeks after you stopped feeding the top of the funnel, you have just run an accidental controlled experiment that tells you what prospecting was actually doing.
Is your branded search volume trending down? Branded search is one of the cleanest signals of brand awareness health. If fewer people are searching for your brand name over time, something upstream of the conversion path is weakening.
Does your CAC look artificially low for branded search and direct? These channels almost always convert in last-click models because they catch customers at the end of a journey that started elsewhere.
What to do about it
Step 1 — Check which attribution model your reports use
In GA4, go to Admin → Attribution settings → Reporting attribution model. GA4 now offers two options: Data-driven and Paid and organic last click. If your property is still on last click, switching to Data-driven changes how conversions are credited in your standard reports.
Note: the model in GA4 does not control how Google Ads reports conversions. Check the attribution model on each conversion action in Google Ads separately.
Step 2 — Build a conversion path report
In GA4, the Advertising → Attribution → Conversion paths report shows you the actual sequences of touchpoints that led to conversions. Filter for multi-step paths. This will tell you which channels appear most frequently as first-touch versus assist versus last-touch.
Step 3 — Run a holdout test before cutting any channel
If you are considering cutting a channel based on last-click performance data, run a geographic or audience holdout test first. Turn off the channel for a segment of customers and measure what happens to overall conversions in that segment over 4–6 weeks. This is the cleanest way to establish whether a channel is genuinely not working or just not getting last-click credit for work it is doing.
Step 4 — Reframe how you evaluate channel performance
Stop using ROAS as the primary metric for top-of-funnel channels. It was designed for direct-response evaluation and it penalises awareness channels that do not convert on first contact. Instead, track branded search volume trend, new user acquisition rate, and assisted conversion count alongside ROAS for a complete picture.
The honest caveat
Multi-touch attribution is better than last-click attribution, but it is not perfect either. Every attribution model makes assumptions about how credit should be divided, and those assumptions are always somewhat arbitrary. The customer journey is too complex and too individual to be perfectly captured by any model.
What multi-touch attribution does is give you a less distorted picture of which channels are contributing to the business — enough to make better budget allocation decisions than you can make with last-click alone.
How much can the model change the picture? In our attribution portfolio project, which compared five models on the same 11,292 touchpoints, Paid Search's share of credit ranged from 14.4% under First Click to 41.9% under Shapley. Same data, very different budget recommendations. That is why any large reallocation should be confirmed with a holdout test before it is scaled.
Key takeaways
- Last-click attribution gives 100% of conversion credit to the final touchpoint before sale, ignoring everything that brought the customer to that point
- It systematically overvalues retargeting and branded search, and undervalues prospecting, awareness and email
- The symptom is retargeting ROAS declining after you cut top-of-funnel spend — the lag gives it away
- Position-based (U-shaped) attribution is a practical starting point for many e-commerce businesses, built from exported path data
- GA4's data-driven model is free and is now the default, but it works best with a healthy volume of conversions
- Run holdout tests before cutting any channel based on last-click performance data alone